Google makes quantum computing’s costliest ingredient 40 times more reliable
A technique called magic state cultivation worked on real hardware. Only 8% of attempts survive, and that is the point.
Known
- Google researchers ran "magic state cultivation" on their Willow quantum processor. - Their paper reports the error fell by a factor of 40, with a fidelity of 0.9999. - Only 8% of attempts were kept.
Not known yet
- How fast these states can be made at the scale a useful quantum computer needs. - When, if ever, this leads to machines that do useful work. The paper does not say.
Quantum computers have a problem that has nothing to do with speed: they make mistakes, constantly. Google researchers have now shown a way to make one of their most expensive ingredients 40 times more reliable.
The Google Quantum AI team ran a technique called "magic state cultivation" on its Willow superconducting processor. Their paper reports that cultivation "reduces the error by a factor of 40, with a state fidelity of 0.9999".
What a magic state is
Ordinary quantum operations can be protected against errors fairly cheaply. But to run useful programs, a quantum computer also needs a special kind of operation that cannot. It gets around this by consuming "magic states", carefully prepared single-use ingredients.
Making good magic states has been one of the most costly parts of the whole design. The usual method, called distillation, takes many noisy states and boils them down to a few good ones.
What Google did
Cultivation, proposed by Google researchers Craig Gidney, Noah Shutty and Cody Jones in 2024, works differently. It prepares a magic state, checks it again and again, and throws it away the moment a check fails.
That is why the result comes with a second number: only 8% of attempts survived. The states that did survive had a fidelity of 0.9999, roughly one error in ten thousand.
The paper calls the result experimental proof that cultivation is "a viable solution to one of quantum computing's most significant challenges."
What it does not mean
This is not a new chip, and it does not make today's quantum computers useful for real work. A useful, error-corrected machine needs huge numbers of these states, produced quickly and joined to many other parts.